AI agents on Microsoft Foundry
TechRam builds AI agents on Microsoft Foundry that run off a trigger, not a prompt. A meeting ends or a status changes, the agent does the job inside your own Microsoft 365 and Azure tenant, and a person approves the result before anything is filed or sent. The records it produces live in SharePoint.
Triggers, not prompts
Most organisations use AI by typing into it. The value shows up when it runs off an event instead, the way an automation does: a meeting ends, a file lands, a project closes. The agent does the work, and a person reviews it before it goes anywhere. That’s how we run our own meetings, and it’s what we build for clients.
Meeting minutes for an ISO 9001 consultancy
A project management consultancy certified to ISO 9001 has to minute its internal meetings. It used to take notes in one AI tool, paste them into another with a Word template, and file the result by hand.
What we built on Foundry:
- Whichever minute-taker or meeting recorder the firm uses passes the transcript to an agent in the firm’s own Azure tenant
- The agent fills in the firm’s own Word minutes template, with a version for each type of meeting
- The process owner reviews and approves the minutes
- They’re filed to the matching SharePoint project site, using the project ID in the meeting title, and anything unmatched lands in a list for someone to sort
Nobody retypes the minutes, and nothing is filed unchecked. The same engagement automated client feedback requests and email filing into SharePoint.
What it needs underneath
An agent can only file to a structure that exists, and it can only answer from what it can reach. That’s why the work starts with The SharePoint Audit, and why agents sit inside AI Workflows & Agents, not in front of it.
Common questions
01. Does the agent send or file anything without a person checking it?
No. The agent drafts, and a named person reviews and approves before anything is filed or sent. That's the rule we design to: the agent does the slow part, and a person still decides what goes out.
02. Where does our data go?
The agent runs in a Foundry project inside your own Azure tenant, reads from your Microsoft 365 and writes back to your SharePoint, and you own the project and the output. If a meeting recorder supplies the transcript, that's the one outside service in the chain, and you choose it.
03. Should we use Copilot Studio or Foundry?
It depends on how the agent is used. If your people will talk to it inside Teams or Copilot, we start in Copilot Studio. If it works on its own off a trigger and runs a multi-step job, like turning a transcript into filed minutes, we build it on Foundry.
04. What does an agent like this cost to build?
A fixed price, not a day rate, agreed in writing before work starts. The first engagement is usually The SharePoint Audit, because an agent can only file to a structure that exists.